Abstract: Objective To explore the relationship between drug quality risk and adverse drug reaction (ADR) and to provide valuable risk warning for regulators through ADR signal mining. Methods A adverse event cluster signals identification model was established and the results of the analysis were visualized. At the same time, retrospective analysis of calcium gluconate injection and citicoline sodium injection using the adverse reaction monitoring data provided by Henan Center was conducted to verify the model. Results The ADR data of Henan enterprises fed back by the National Center from March 1, 2015 to April 4, 2015 and the data collected by Henan Center from March 20, 2014 to April 9, 2014 were analyzed by the cluster signals identification model, respectively. Adverse reactions data were tested for one cycle for 7 days. Risk signals of lot number 14102421 of YSZY calcium gluconate injection were detected in the fourth cycle of the first period .The risk signals of AHLYYY's citicoline 130727 was detected in the second cycle of second period. Results of model analysis were in line with the situation of quality sampling. Conclusion The Methods of cluster signals detection and identification designed in this study are feasible and can be used for initial screening of specific batch risk of enterprise varieties. Taking into account only the use of ADR monitoring data in Henan province, as well as a small number of cases retrospective validation, the applicability of the model still needs further verification. In addition, this study uses visual processing to make the output of the model more intuitive, meets the actual needs of regulation and provides a reference for risk-based regulation, and we will continue to improve the system in the next step.

Key words: drug quality risk, supervision suggestion, visualization method, ADR data mining, cluster risk signals